メニュー
Use when integrating a new source (article, paper, note, meeting transcript, book chapter) into an LLM Wiki — reads the source, writes a summary page, updates entity/concept/topic pages, maintains cross-references, and records the work in the index and log.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when setting up a new personal LLM Wiki — an agent-maintained, interlinked markdown knowledge base. Interviews the user about their domain, scaffolds the raw/wiki/schema structure, writes a tailored schema file (CLAUDE.md / AGENTS.md), and seeds the index, log, and overview.
Use when health-checking an LLM Wiki — finds contradictions, stale claims, orphan pages, broken links, missing cross-references, and schema violations, then produces a prioritized report and offers fixes.
Use when answering a question from an LLM Wiki — navigates the index, reads relevant pages, synthesizes a cited answer, and offers to file valuable answers back into the wiki as analysis pages so explorations compound.
Use when the user wants Tiro notes to flow into their LLM Wiki automatically on a schedule — registers a recurring job (agent scheduled task or cron) that runs the sync workflow daily or at a chosen cadence.
Use when the user wants to deeply learn a concept, form an opinion, or stress-test their understanding through fact-grounded Socratic dialogue — a 50:50 mix of providing information and challenging with counterarguments, grounded in their wiki and web research.
Use when syncing Tiro meeting notes and transcripts into an LLM Wiki — discovers notes not yet ingested, downloads them into raw/ via the Tiro CLI, and runs the ingest workflow on each, idempotently and without ever modifying existing raw files.
| name | ingest |
| description | Use when integrating a new source (article, paper, note, meeting transcript, book chapter) into an LLM Wiki — reads the source, writes a summary page, updates entity/concept/topic pages, maintains cross-references, and records the work in the index and log. |
Take one source from raw/ and compile it into the wiki: summary page, entity/concept updates, cross-references, index, log.
/tiro:ingest [file path, URL, or description]
If no argument is given, look for files added to raw/ in the last 24 hours and let the user pick.
raw/ subfolder first — raw/ is the source of truth, the wiki never links to content that could disappear.wiki/sources/:
source_path (back-reference to the raw file), plus title, type, date, tags per the wiki's schema.index.md to see what exists, then:
index.md, append a log entry (## [YYYY-MM-DD] ingest | Title with created/modified page lists), refresh overview.md counts.For many sources at once: create source pages in parallel if you use subagents, but never let two writers touch shared files (index.md, log.md, entity/concept/topic pages). Merge those sequentially at the end.